HARO Pilot

AI-Powered HARO Response System

Open dashboard

Inbound HARO emails come in messy. The review layer should not.

This build is structured around a simple persisted workflow: store the email, extract multiple opportunities, classify them, run conditional research, draft across three providers, then require a human edit before anything goes out.

Workflow

  1. 1. Webhook verifies and stores raw inbound email.
  2. 2. State machine advances extraction, classification, and research.
  3. 3. Grok, Claude, and Gemini produce parallel drafts.
  4. 4. User edits the final response before send.

Multiple opportunities per email

The schema and workflow assume one inbound email can yield many requests.

No queue theater

Progress is driven by explicit database statuses and one processor route.

Manual control before send

The final draft is always editable before approval.